Torras.ai

How We Got Here—and Where AI Is Taking Us Next

Earlier this month, I had the pleasure of speaking at an event hosted by Ingecal, as they celebrated an impressive 25 years of helping organizations implement management systems and improve their processes. I was honored to contribute to such a meaningful occasion and share insights into the ever-evolving world of artificial intelligence.

For those who don’t know me, my name is Jordi Torras. I’ve been working with AI since the 1980s, back when we didn’t even call it “AI” the way we do now. I started as a computer science student when it was still a strange, new field. Together with my brother, I co-founded a company called SBD, which we eventually sold. Later, I founded another company focused entirely on AI, moved to the San Francisco Bay Area, raised venture capital, and built products powered by intelligent algorithms long before it became trendy.

We often think of AI as futuristic, but it’s already here—and it has been here for a long time. What’s changed is its power, its speed, and how seamlessly it’s becoming part of our everyday lives.

The Journey from ISO 9000 to ISO 27001

One story I shared during my talk was about our first brush with ISO 9000 certification. At the time, we were a small company, and it felt like something only big, corporate organizations dealt with. I remember googling what ISO 9000 even was (there was no ChatGPT back then!), and feeling completely overwhelmed. But as with many things in life, we found the right people—like Evaristo Gutiérrez —and followed the process. It wasn’t magic. It was method. And the result? We got certified. Years later, with my next company—this one in the AI and cloud software space—we had to go through the same thing with ISO 27001, focused on data security. Again, not because we wanted to, but because we had to. And again, we succeeded.

From Fictional Robots to Neural Networks

When most people think of AI, they imagine robots. Characters from Star Wars, or the Terminator, or HAL 9000 from 2001: A Space Odyssey. That last one is actually my favorite. HAL wasn’t a robot with legs and arms. It was a voice, a camera lens, a system—just like today’s AI.

But real AI doesn’t look like a movie. It started with a much more humble inspiration: the human brain. The real brain doesn’t have gears or circuits. It’s made of neurons—billions of them—that pass electrical signals around. And that concept gave rise to what we now call artificial neural networks. Each artificial neuron does a simple math operation, but when you link them all together in layers, suddenly you get something powerful. Something that can learn.

The Magic of Learning From Data

I showed an example during the talk of a neural network that learns to recognize handwritten digits—0 to 9. It doesn’t know what a “2” is. It’s just been trained on thousands of examples. It sees pixels, not numbers. And after enough examples, it gets really good at saying, “That’s a 7,” or “That’s a 3.” That’s machine learning. No magic. Just lots of data, and lots of pattern recognition.

ChatGPT works the same way, just on a much larger scale. Instead of recognizing numbers, it recognizes language patterns. It’s been trained on what is essentially the entire internet—Wikipedia, blog posts, news articles, forums (yes, even ForoCoches!). It doesn’t understand language the way we do. But it knows what word is likely to come next, based on what it’s seen before. That’s it. It’s a probability machine, predicting the next word, then the next, and so on. And yet, the results are astonishing.

Why ChatGPT Changed the Game

ChatGPT didn’t just go viral—it exploded. It reached a million users in five days. For comparison, Netflix took 3.5 years, and Spotify five months. ChatGPT needed no advertising. Just word of mouth. People saw it, used it, and instantly recognized the value.

But this level of AI doesn’t come cheap. Training a large language model like GPT-4 costs tens or even hundreds of millions of dollars. That’s why companies like NVIDIA, who make the specialized GPUs required for training, have seen their stock value explode. If you had invested back in 2000… well, let’s just say it would’ve been a very good investment.

The Pyramid of AI

I often describe the evolution of AI as a pyramid. At the bottom, you have basic computation—like a calculator. Then automation, where software replaces repetitive tasks. Above that, you get narrow AI—systems that can do one thing really well, like Google Maps or facial recognition. And at the very top? General AI. A single system that can do anything a human can do: drive a car, play chess, write poetry, make scientific discoveries.

We’re not there yet. But the moment we cross that line—if we ever do—the jump to superintelligence might happen in milliseconds. Machines smarter than us. A future where we’re no longer the smartest beings in the room. That will be a fascinating moment in human history.

Final Thoughts

AI is not science fiction. It’s science fact. And the journey from neurons to neural networks, from calculators to ChatGPT, is just beginning. We live in a world where intelligence is no longer just biological. It’s digital. And what we do with that intelligence—how we apply it, regulate it, and share it—will define the future.

Thanks again to Ingecal for the opportunity to speak, and to everyone who attended or is watching the video now. The conversation is just starting, and I’m excited to be part of it.